GLM-5.3 didn’t change the base model — where did its coding gains come from?

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

Z.ai released GLM-5.3 on Friday, a coding and agent model built from the same base model as GLM-5.2. Developers can already use GLM-5.3 through Z.ai’s GLM Coding Plan with Claude Code, Cline, OpenCode and Codex. Direct API access is still listed as “coming soon,” because Z.ai plans to release the model weights after two weeks of hardening and safety testing.

Post-training did the heavy lifting

Z.ai significantly expanded post-training for GLM-5.3, exposing the model to tenfold more long-horizon task environments while broadening its access to developer tools and engineering workflows.

Some training tasks simulated the full software lifecycle — from identifying bugs and drafting fixes to writing code, running tests and shipping results. According to Z.ai, single tasks matched the workload of a senior engineer over several days.

Z.ai concentrated compute on the specific environments the model actually works in.

This makes GLM-5.3 a compelling case study for post-training compute scaling. Z.ai concentrated compute on the specific environments the model actually works in. They are not alone in testing that hypothesis — DeepSeek recently proved a smaller model could outperform its flagship counterpart simply by optimizing post-training rather than inflating parameter counts.

Benchmarks jump, caveats remain

Z.ai claims GLM-5.3 delivers a 50% performance boost over GLM-5.2 on its internal Code Bench, though self-reported vendor numbers warrant the usual skepticism until the community tests the released weights. On public evaluations, however, the model posts dramatic gains …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://thenewstack.io/glm-5-3-post-training-coding/
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.